Defect Analysis in Compressor Vanes Using Split Bregman Noise-Reduced Digital Industrial Radiography

• Published in: Russian journal of nondestructive testing Vol. 59; no. 5; pp. 622 - 632
• Main Authors: Movafeghi, Amir, Yahaghi, Effat, Monem, Sajad, Nekouei, Jahangir, Rostami, Peyman, Mirzapour, Mahdi
• Format: Journal Article
• Language: English
• Published: Moscow Pleiades Publishing 01-05-2023; Springer Nature B.V
• Subjects: Algorithms; Blurring; Centrifugal compressors; Characterization and Evaluation of Materials; Chemistry and Materials Science; Defects; Digital imaging; Fogging; High speed; Image acquisition; Image quality; Information retrieval; Inspection; Materials Science; Noise reduction; Nondestructive testing; Preventive maintenance; Quality control; Radiography; Structural Materials; Vanes; X-Ray Methods; split Bregman method; compressor vanes; digital industrial radiography; defect analysis
• DOI: 10.1134/S1061830923600211
• EISSN: 1608-3385
• ISSN: 1061-8309

Detection and analysis of defects in the vanes of centrifugal compressors are invaluable as part of routine quality control and preventative maintenance procedures. Historically, radiography testing (RT) and more recently digital industrial radiography testing (DIR) has proved to be an indispensable nondestructive inspection method capable of providing high defect detection sensitivity. Radiographic images are intrinsically noisy and various digital image acquisition and processing methods have been employed to achieve improved selective region and bandwidth image quality. The noise in the RT images is usually random in nature and the quality of the image is often uniformly degraded by the resulting fogging effect. In this study, a modified form of the split Bregman (SB) algorithm was developed and applied to remove the blurring in the radiography images. The SB algorithm was successfully applied to radiographic images of compressor vanes to achieve noise reduction whilst retaining sufficient image details, enabling more effective information extraction of object structures and defects. The results of the study have shown marked and quantifiable improvement in defect detectability and analysis compared to the outcomes using original unprocessed images. The method was used successfully by industrial radiography experts to detect object deterioration due to pre-failure high-speed rotational loading of the vanes; the experts' preference for working with noise-subtracted images was demonstrated as part of the study. These evaluations were carried out after and before high-speed testing of the vanes. It was confirmed that the implemented method improved image quality and defect detection.